03_Redemption_and_Expected_Spend_Prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3fcc5adc",
   "metadata": {},
   "source": [
    "# Improving Starbucks Offer Targeting with Redemption Prediction\n",
    "\n",
    "## 03. Redemption and Expected Spend Prediction\n",
    "\n",
    "This is the third & final notebook out of 3 which details the steps taken to:\n",
    "1. Tune and select a classifier that predicts redemption\n",
    "2. Tuning and selecting a regressor that predicts expected spend if redeemed \n",
    "3. Combine both to output an understandable ranking of offers\n",
    "\n",
    "<b> Content </b>\n",
    "- [0. Recap of Variables](#section0)\n",
    "- [1. Prepare Dataset](#section1)\n",
    "    - [1.1 Encode Categorical Variables](#section1.1)\n",
    "- [2. Predicting Offer Redemption through Classfication](#section2)\n",
    "    - [2.1 Create Train-Test Split & Evaluation Helper Functions](#section2.1)\n",
    "    - [2.2 Logistic Regression](#section2.2)\n",
    "    - [2.3 Support Vector Machine](#section2.3)\n",
    "    - [2.4 Random Forest Classifier](#section2.4)\n",
    "    - [2.5 XGBoost Classifier](#section2.5)\n",
    "    - [2.6 Classifier Evaluation & Selection](#section2.6)\n",
    "    - [2.7 Findings: The Most Important Features Predicting Offer Success](#section2.7)\n",
    "- [3. Predicting Amount Spent through Regression](#section3)\n",
    "    - [3.1 Create Train-Test Split and Evaluation Functions](#section3.1)\n",
    "    - [3.2 Linear Regression](#section3.2)\n",
    "    - [3.3 Random Forest Regressor](#section3.3)\n",
    "    - [3.4 K-Nearest Neighbours Regressor](#section3.4)\n",
    "    - [3.5 Regressor Evaluation & Selection](#section3.5)\n",
    "    - [3.6 Findings: The Most Important Features Predicting Amount Spent](#section3.6)\n",
    "- [4. Combined Expected Value Model](#section4)\n",
    "    - [4.1 Offer Ranking Using Trained Models](#section4.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "69c076bd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import numerical libraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "# Import plotting libraries\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Import pre-processing functions\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "# Import tuning functions\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "\n",
    "# Import classifiers & regressors\n",
    "from sklearn.linear_model import LogisticRegression, LinearRegression\n",
    "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "from sklearn.svm import SVC\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "# Import evaluation functions\n",
    "from sklearn.metrics import accuracy_score, confusion_matrix, precision_score, \\\n",
    "                            ConfusionMatrixDisplay,roc_auc_score, roc_curve, recall_score,mean_squared_error \n",
    "from sklearn.datasets import make_circles\n",
    "from sklearn.inspection import permutation_importance\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75a1833f",
   "metadata": {},
   "source": [
    "## 0. Recap of Variables<a class=\"anchor\" id=\"section0\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe7175e3",
   "metadata": {},
   "source": [
    "This data was already processed in Notebook *01.Data Preprocessing*.\n",
    "\n",
    "- *person*: Alphanumeric identifier of a customer\n",
    "- *offer_id*: Alphanumeric identifier of an offer\n",
    "- *journey*: Alphanumeric identifier of an end-to-end offer lifecycle\n",
    "- *days*: Days since data collection has commenced\n",
    "- *offer received, viewed, completed*: 0 if True, 1 if False\n",
    "- *gender*: Male (M), Female (F) or Others (O)\n",
    "- *age*: Numeric value for age\n",
    "- *reward*: The number of reward points (stars*) awarded for completion of an offer\n",
    "- *difficulty*: Minimum spend required to complete offer\n",
    "- *duration*: Days that offer is valid for\n",
    "- *offer type*: Buy-one-get-one (bogo), informational or direct discount\n",
    "- *channels*: Communication channels that offer was delivered over, split into dummy \n",
    "- *promo_name*: Promotion configuration identifier\n",
    "- *amount*: Transaction amount\n",
    "- *catcoding*: Combination of offer received-offer viewed-offer completed as per the table below\n",
    "- *scenario_description*: User-friendly description of *catcoding* as per the table below\n",
    "\n",
    "<img src = \"./images/customerjourneywstatus.png\" width=\"700\"/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "d33cf91b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['person', 'offer_id', 'journey', 'days', 'offer received',\n",
       "       'offer viewed', 'offer completed', 'gender', 'age', 'income',\n",
       "       'member_duration', 'reward', 'difficulty', 'duration', 'offer_type',\n",
       "       'channel_web', 'channel_mobile', 'channel_social', 'promo_name',\n",
       "       'amount', 'catcoding', 'scenario_description'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Import and view data\n",
    "df_all = pd.read_csv('data/preprocessed_data.csv')\n",
    "df_all.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "17bd33d4",
   "metadata": {},
   "source": [
    "## 1. Prepare Dataset<a class=\"anchor\" id=\"section1\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3b8cd22f",
   "metadata": {},
   "source": [
    "### 1.1 Encode Categorical Variables <a class=\"anchor\" id=\"section1.1\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed3f1669",
   "metadata": {},
   "source": [
    "First, create a dummy column with the classification target variable *Success*. If offer was received, seen and completed, then *Success* = 1, else 0."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "2757d055",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Add a binary success / failure column\n",
    "df_all['success'] = np.where(df_all[\"scenario_description\"] == 'success',1,0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "acbf9e1f",
   "metadata": {},
   "source": [
    "Secondly, create dummies for categorical columns.\n",
    "\n",
    "- *gender_dum_M*: 0 if Female, 1 if Male\n",
    "- *offer_type_dum_discount*: 0 if buy-one-get-one, 1 if discount"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "2bde66fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Copy the dummy columns to preserve them\n",
    "df_all[['offer_type_dum', 'gender_dum']] = df_all[['offer_type','gender']]\n",
    "\n",
    "# Use newly created columns in get_dummies (original columns will be dropped)\n",
    "dummies_needed = ['gender_dum', 'offer_type_dum']\n",
    "\n",
    "# Ensure to drop_first = True to prevent multicollinearity issues\n",
    "df_all = pd.get_dummies(data=df_all, columns=dummies_needed, drop_first=True, prefix_sep='_')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "31d9b736",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>person</th>\n",
       "      <th>offer_id</th>\n",
       "      <th>journey</th>\n",
       "      <th>days</th>\n",
       "      <th>offer received</th>\n",
       "      <th>offer viewed</th>\n",
       "      <th>offer completed</th>\n",
       "      <th>gender</th>\n",
       "      <th>age</th>\n",
       "      <th>income</th>\n",
       "      <th>...</th>\n",
       "      <th>channel_web</th>\n",
       "      <th>channel_mobile</th>\n",
       "      <th>channel_social</th>\n",
       "      <th>promo_name</th>\n",
       "      <th>amount</th>\n",
       "      <th>catcoding</th>\n",
       "      <th>scenario_description</th>\n",
       "      <th>success</th>\n",
       "      <th>gender_dum_M</th>\n",
       "      <th>offer_type_dum_discount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0009655768c64bdeb2e877511632db8f</td>\n",
       "      <td>2906b810c7d4411798c6938adc9daaa5</td>\n",
       "      <td>1</td>\n",
       "      <td>24.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>M</td>\n",
       "      <td>33</td>\n",
       "      <td>72000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>disc-1</td>\n",
       "      <td>10.27</td>\n",
       "      <td>1-0-1</td>\n",
       "      <td>failed, completed</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>00426fe3ffde4c6b9cb9ad6d077a13ea</td>\n",
       "      <td>2906b810c7d4411798c6938adc9daaa5</td>\n",
       "      <td>30</td>\n",
       "      <td>17.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>F</td>\n",
       "      <td>19</td>\n",
       "      <td>65000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>disc-1</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1-0-0</td>\n",
       "      <td>failed, not seen</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>00426fe3ffde4c6b9cb9ad6d077a13ea</td>\n",
       "      <td>2906b810c7d4411798c6938adc9daaa5</td>\n",
       "      <td>31</td>\n",
       "      <td>24.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>F</td>\n",
       "      <td>19</td>\n",
       "      <td>65000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>disc-1</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1-0-0</td>\n",
       "      <td>failed, not seen</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>005500a7188546ff8a767329a2f7c76a</td>\n",
       "      <td>2906b810c7d4411798c6938adc9daaa5</td>\n",
       "      <td>43</td>\n",
       "      <td>17.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>M</td>\n",
       "      <td>56</td>\n",
       "      <td>47000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>disc-1</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1-0-0</td>\n",
       "      <td>failed, not seen</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>00715b6e55c3431cb56ff7307eb19675</td>\n",
       "      <td>2906b810c7d4411798c6938adc9daaa5</td>\n",
       "      <td>65</td>\n",
       "      <td>22.25</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>F</td>\n",
       "      <td>58</td>\n",
       "      <td>119000.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>disc-1</td>\n",
       "      <td>21.35</td>\n",
       "      <td>1-0-1</td>\n",
       "      <td>failed, completed</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                             person                          offer_id  \\\n",
       "0  0009655768c64bdeb2e877511632db8f  2906b810c7d4411798c6938adc9daaa5   \n",
       "1  00426fe3ffde4c6b9cb9ad6d077a13ea  2906b810c7d4411798c6938adc9daaa5   \n",
       "2  00426fe3ffde4c6b9cb9ad6d077a13ea  2906b810c7d4411798c6938adc9daaa5   \n",
       "3  005500a7188546ff8a767329a2f7c76a  2906b810c7d4411798c6938adc9daaa5   \n",
       "4  00715b6e55c3431cb56ff7307eb19675  2906b810c7d4411798c6938adc9daaa5   \n",
       "\n",
       "   journey   days  offer received  offer viewed  offer completed gender  age  \\\n",
       "0        1  24.00               1             0                1      M   33   \n",
       "1       30  17.00               1             0                0      F   19   \n",
       "2       31  24.00               1             0                0      F   19   \n",
       "3       43  17.00               1             0                0      M   56   \n",
       "4       65  22.25               1             0                1      F   58   \n",
       "\n",
       "     income  ...  channel_web  channel_mobile  channel_social  promo_name  \\\n",
       "0   72000.0  ...            1               1               0      disc-1   \n",
       "1   65000.0  ...            1               1               0      disc-1   \n",
       "2   65000.0  ...            1               1               0      disc-1   \n",
       "3   47000.0  ...            1               1               0      disc-1   \n",
       "4  119000.0  ...            1               1               0      disc-1   \n",
       "\n",
       "  amount  catcoding  scenario_description  success gender_dum_M  \\\n",
       "0  10.27      1-0-1     failed, completed        0            1   \n",
       "1   0.00      1-0-0      failed, not seen        0            0   \n",
       "2   0.00      1-0-0      failed, not seen        0            0   \n",
       "3   0.00      1-0-0      failed, not seen        0            1   \n",
       "4  21.35      1-0-1     failed, completed        0            0   \n",
       "\n",
       "   offer_type_dum_discount  \n",
       "0                        1  \n",
       "1                        1  \n",
       "2                        1  \n",
       "3                        1  \n",
       "4                        1  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_all.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99518552",
   "metadata": {},
   "source": [
    "## 2. Predicting Offer Redemption through Classfication<a class=\"anchor\" id=\"section2\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a4f7264f",
   "metadata": {},
   "source": [
    "#### Target Variable: *success* (binary)\n",
    "\n",
    "In this section, we train and select a classifier that helps to predict offer success. This will help to identify which offers are most likely to be redeemed by a particular customer."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1cec13cf",
   "metadata": {},
   "source": [
    "### 2.1 Create Train-Test Split & Evaluation Helper Functions<a class=\"anchor\" id=\"section2.1\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "714fdd4b",
   "metadata": {},
   "source": [
    "30% of data is withheld as the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "91dd7dbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "y = df_all['success']\n",
    "\n",
    "features = ['age', 'income', 'member_duration','reward', 'difficulty', \n",
    "        'days', 'channel_web', 'channel_mobile', \n",
    "        'channel_social', 'gender_dum_M', 'offer_type_dum_discount']\n",
    "\n",
    "X = df_all[features]\n",
    "\n",
    "# Define train-test split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3,\n",
    "                                                    random_state=42, stratify=y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "4ccb73fd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a comprehensive classification summary for easy checking\n",
    "\n",
    "def classification_summary(X_train, X_test, y_train, y_test, model_name, model_name_text):\n",
    "    \n",
    "    y_pred_test = model_name.predict(X_test)\n",
    "    \n",
    "    cm = confusion_matrix(y_test, y_pred_test, labels=model_name.classes_)\n",
    "\n",
    "    # Compute all metrics manually\n",
    "    tn = cm[0][0]\n",
    "    fp = cm[0][1]\n",
    "    fn = cm[1][0]\n",
    "    tp = cm[1][1]\n",
    "\n",
    "    tpr = tp/(tp+fn)\n",
    "    fnr = 1 - tpr\n",
    "\n",
    "    tnr = tn/(tn+fp)\n",
    "    fpr = 1-tnr\n",
    "\n",
    "    print('Test Set Metrics')\n",
    "    print('{} Plain Accuracy:{}%'.format(model_name_text,\n",
    "                                         round(accuracy_score(y_test, y_pred_test)*100,2)))\n",
    "    print('TPR:{}%'.format(round(tpr*100,2)))\n",
    "    print('FNR:{}%'.format(round(fnr*100,2)))\n",
    "    print('TNR:{}%'.format(round(tnr*100,2)))\n",
    "    print('FPR:{}%'.format(round(fpr*100,2)))\n",
    "    print('Recall:{}%'.format(round(recall_score(y_test, y_pred_test)*100,2)))\n",
    "    print('Precision:{}%'.format(round(precision_score(y_test, y_pred_test)*100,2)))\n",
    "    print('AUC Score:{}'.format(round(roc_auc_score(y_test,y_pred_test),3)))\n",
    "    print('')\n",
    "    \n",
    "    # Compute train set accuracy\n",
    "    print('Train Set Metrics')\n",
    "    y_pred_train = model_name.predict(X_train)\n",
    "    print('{} Plain Accuracy:{}%'.format(model_name_text,\n",
    "                                         round(accuracy_score(y_train, y_pred_train)*100,2)))\n",
    "    print('Recall:{}%'.format(round(recall_score(y_train, y_pred_train)*100,2)))\n",
    "    print('Precision:{}%'.format(round(precision_score(y_train, y_pred_train)*100,2)))\n",
    "    print('AUC Score:{}'.format(round(roc_auc_score(y_train,y_pred_train),3)))    \n",
    "    \n",
    "    # Display confusion matrix\n",
    "    disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=model_name.classes_)\n",
    "    disp.plot()\n",
    "    plt.show()\n",
    "    \n",
    "    # Compute predicted probabilities: y_pred_prob\n",
    "    y_pred_prob = model_name.predict_proba(X_test)[:,1]\n",
    "\n",
    "    # Generate ROC curve values: fpr, tpr, thresholds\n",
    "    fpr, tpr, thresholds = roc_curve(y_test, y_pred_prob)\n",
    "\n",
    "    # Plot ROC curve\n",
    "    fig, ax = plt.subplots(figsize=[8,6])\n",
    "    plt.plot([0, 1], [0, 1], 'k--')\n",
    "    plt.plot(fpr, tpr)\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.title('ROC Curve ({})'.format(model_name_text), fontsize = 20)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "10bf2c04",
   "metadata": {},
   "source": [
    "### 2.2 Logistic Regression<a class=\"anchor\" id=\"section2.2\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "711135c5",
   "metadata": {},
   "source": [
    "#### 2.2.1 Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "042b7c8f",
   "metadata": {},
   "source": [
    "Hyperparameters are tuned using 5-fold cross validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "a8b1daad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'C': 10, 'penalty': 'l1', 'solver': 'liblinear'}\n",
      "Best score is 0.6737011155502943\n"
     ]
    }
   ],
   "source": [
    "param_grid = {\n",
    "     'penalty' : ['l1', 'l2'],\n",
    "    'C' : [0.001, 0.1, 1, 5, 10, 20, 50], #default = 1\n",
    "    'solver' : ['liblinear']}\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = LogisticRegression()\n",
    "lr_cv = GridSearchCV(model, param_grid, cv=5)\n",
    "lr_cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(lr_cv.best_params_)) \n",
    "print(\"Best score is {}\".format(lr_cv.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "786b5f70",
   "metadata": {},
   "source": [
    "#### 2.2.2 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7fad33f2",
   "metadata": {},
   "source": [
    "Pass the best parameters into Logistic Regression model and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "d4ce470b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Metrics\n",
      "Logistic Regression Plain Accuracy:67.19%\n",
      "TPR:51.46%\n",
      "FNR:48.54%\n",
      "TNR:78.56%\n",
      "FPR:21.44%\n",
      "Recall:51.46%\n",
      "Precision:63.44%\n",
      "AUC Score:0.65\n",
      "\n",
      "Train Set Metrics\n",
      "Logistic Regression Plain Accuracy:67.42%\n",
      "Recall:52.53%\n",
      "Precision:63.49%\n",
      "AUC Score:0.654\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "logreg = LogisticRegression(C = 10, penalty = 'l1', solver = 'liblinear')\n",
    "logreg.fit(X_train, y_train)\n",
    "\n",
    "classification_summary(X_train = X_train, \n",
    "                       X_test = X_test,\n",
    "                       y_train = y_train, \n",
    "                       y_test = y_test, \n",
    "                       model_name = logreg, \n",
    "                       model_name_text = 'Logistic Regression')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ea18d9d6",
   "metadata": {},
   "source": [
    "### 2.3 Support Vector Machine<a class=\"anchor\" id=\"section2.3\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25df2c43",
   "metadata": {},
   "source": [
    "#### 2.3.1 Perform Feature Scaling\n",
    "\n",
    "Features must first be scaled because having inconsistent scales for different features distorts the closeness of each observation, with some features having more influence than others.\n",
    "\n",
    "- Binary variables are not scaled\n",
    "- After scaling, scaled variables have mean of 0 and standard deviation of 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4f3315c7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>age</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-2.17</td>\n",
       "      <td>-0.67</td>\n",
       "      <td>0.08</td>\n",
       "      <td>0.76</td>\n",
       "      <td>1.89</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>income</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.63</td>\n",
       "      <td>-0.75</td>\n",
       "      <td>-0.10</td>\n",
       "      <td>0.64</td>\n",
       "      <td>2.53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>member_duration</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.24</td>\n",
       "      <td>-0.75</td>\n",
       "      <td>-0.40</td>\n",
       "      <td>0.66</td>\n",
       "      <td>3.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>reward</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.09</td>\n",
       "      <td>-1.09</td>\n",
       "      <td>-0.08</td>\n",
       "      <td>1.58</td>\n",
       "      <td>1.58</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>difficulty</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.05</td>\n",
       "      <td>-0.60</td>\n",
       "      <td>0.08</td>\n",
       "      <td>0.08</td>\n",
       "      <td>2.34</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>days</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.89</td>\n",
       "      <td>-0.89</td>\n",
       "      <td>0.19</td>\n",
       "      <td>0.82</td>\n",
       "      <td>1.69</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_web</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.87</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_mobile</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.88</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_social</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.63</td>\n",
       "      <td>0.48</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gender_dum_M</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.59</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>offer_type_dum_discount</th>\n",
       "      <td>34876.0</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           count  mean   std   min   25%   50%   75%   max\n",
       "age                      34876.0 -0.00  1.00 -2.17 -0.67  0.08  0.76  1.89\n",
       "income                   34876.0  0.00  1.00 -1.63 -0.75 -0.10  0.64  2.53\n",
       "member_duration          34876.0 -0.00  1.00 -1.24 -0.75 -0.40  0.66  3.10\n",
       "reward                   34876.0  0.00  1.00 -1.09 -1.09 -0.08  1.58  1.58\n",
       "difficulty               34876.0  0.00  1.00 -1.05 -0.60  0.08  0.08  2.34\n",
       "days                     34876.0  0.00  1.00 -1.89 -0.89  0.19  0.82  1.69\n",
       "channel_web              34876.0  0.87  0.33  0.00  1.00  1.00  1.00  1.00\n",
       "channel_mobile           34876.0  0.88  0.33  0.00  1.00  1.00  1.00  1.00\n",
       "channel_social           34876.0  0.63  0.48  0.00  0.00  1.00  1.00  1.00\n",
       "gender_dum_M             34876.0  0.59  0.49  0.00  0.00  1.00  1.00  1.00\n",
       "offer_type_dum_discount  34876.0  0.50  0.50  0.00  0.00  1.00  1.00  1.00"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Create a ColumnTransformer helper\n",
    "scale_cols = ['age', 'income', 'member_duration', 'reward', 'difficulty', 'days']\n",
    "ct = ColumnTransformer([('transformer', StandardScaler(), scale_cols)], \n",
    "                       remainder='passthrough')\n",
    "\n",
    "\n",
    "# Scale the Train Set and check summary statistics\n",
    "X_train_scaled = X_train.copy()\n",
    "X_train_scaled = pd.DataFrame(ct.fit_transform(X_train_scaled), \n",
    "                              columns = features) # features used to define X_train\n",
    "X_train_scaled.describe().transpose().round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "23421e48",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>age</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-2.16</td>\n",
       "      <td>-0.73</td>\n",
       "      <td>0.09</td>\n",
       "      <td>0.77</td>\n",
       "      <td>1.90</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>income</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.63</td>\n",
       "      <td>-0.75</td>\n",
       "      <td>-0.10</td>\n",
       "      <td>0.65</td>\n",
       "      <td>2.56</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>member_duration</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.24</td>\n",
       "      <td>-0.75</td>\n",
       "      <td>-0.39</td>\n",
       "      <td>0.65</td>\n",
       "      <td>3.07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>reward</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.09</td>\n",
       "      <td>-0.76</td>\n",
       "      <td>-0.08</td>\n",
       "      <td>-0.08</td>\n",
       "      <td>1.61</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>difficulty</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.03</td>\n",
       "      <td>-1.03</td>\n",
       "      <td>0.08</td>\n",
       "      <td>0.08</td>\n",
       "      <td>2.30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>days</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>-0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>-1.91</td>\n",
       "      <td>-0.88</td>\n",
       "      <td>0.18</td>\n",
       "      <td>0.85</td>\n",
       "      <td>1.70</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_web</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.87</td>\n",
       "      <td>0.33</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_mobile</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.87</td>\n",
       "      <td>0.34</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>channel_social</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.62</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gender_dum_M</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.59</td>\n",
       "      <td>0.49</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>offer_type_dum_discount</th>\n",
       "      <td>14948.0</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.50</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           count  mean   std   min   25%   50%   75%   max\n",
       "age                      14948.0  0.00  1.00 -2.16 -0.73  0.09  0.77  1.90\n",
       "income                   14948.0 -0.00  1.00 -1.63 -0.75 -0.10  0.65  2.56\n",
       "member_duration          14948.0 -0.00  1.00 -1.24 -0.75 -0.39  0.65  3.07\n",
       "reward                   14948.0 -0.00  1.00 -1.09 -0.76 -0.08 -0.08  1.61\n",
       "difficulty               14948.0  0.00  1.00 -1.03 -1.03  0.08  0.08  2.30\n",
       "days                     14948.0 -0.00  1.00 -1.91 -0.88  0.18  0.85  1.70\n",
       "channel_web              14948.0  0.87  0.33  0.00  1.00  1.00  1.00  1.00\n",
       "channel_mobile           14948.0  0.87  0.34  0.00  1.00  1.00  1.00  1.00\n",
       "channel_social           14948.0  0.62  0.49  0.00  0.00  1.00  1.00  1.00\n",
       "gender_dum_M             14948.0  0.59  0.49  0.00  0.00  1.00  1.00  1.00\n",
       "offer_type_dum_discount  14948.0  0.50  0.50  0.00  0.00  1.00  1.00  1.00"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Scale the Test Set and check summary statistics\n",
    "X_test_scaled = X_test.copy()\n",
    "X_test_scaled = pd.DataFrame(ct.fit_transform(X_test_scaled), columns = features)\n",
    "X_test_scaled.describe().transpose().round(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0a64060",
   "metadata": {},
   "source": [
    "#### 2.3.2 Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ecec24d",
   "metadata": {},
   "source": [
    "Hyperparameters are tuned using 5-fold cross validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "b4286554",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'C': 10, 'kernel': 'rbf'}\n",
      "Best score is 0.7048686700535991\n"
     ]
    }
   ],
   "source": [
    "# Kernel to be tuned in the following\n",
    "C = [0.1, 1, 10, 50]\n",
    "kernel = ['poly','rbf','sigmoid']\n",
    "\n",
    "param_grid = {'C': C, #default = 1\n",
    "              'kernel': kernel} #default = RBF\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = SVC(random_state = 42) # default kernel = rbf\n",
    "cv = GridSearchCV(model, param_grid, cv=5, n_jobs = -1)\n",
    "cv.fit(X_train_scaled,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(cv.best_params_)) \n",
    "print(\"Best score is {}\".format(cv.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "93901619",
   "metadata": {},
   "source": [
    "#### 2.3.3 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2231b433",
   "metadata": {},
   "source": [
    "Pass the best parameters into Support Vector Classifier and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "89993755",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Metrics\n",
      "Support Vector Machine Plain Accuracy:70.23%\n",
      "TPR:54.49%\n",
      "FNR:45.51%\n",
      "TNR:81.61%\n",
      "FPR:18.39%\n",
      "Recall:54.49%\n",
      "Precision:68.16%\n",
      "AUC Score:0.68\n",
      "\n",
      "Train Set Metrics\n",
      "Support Vector Machine Plain Accuracy:71.21%\n",
      "Recall:56.06%\n",
      "Precision:69.43%\n",
      "AUC Score:0.691\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "svm = SVC(probability = True, C = 10, gamma = 'scale', kernel = 'rbf')\n",
    "svm.fit(X_train_scaled, y_train)\n",
    "\n",
    "classification_summary(X_train = X_train_scaled, \n",
    "                       X_test = X_test_scaled,\n",
    "                       y_train = y_train, \n",
    "                       y_test = y_test, \n",
    "                       model_name = svm, \n",
    "                       model_name_text = 'Support Vector Machine')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05d9b7ad",
   "metadata": {},
   "source": [
    "### 2.4 Random Forest Classifier<a class=\"anchor\" id=\"section2.4\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5ac15cd1",
   "metadata": {},
   "source": [
    "#### 2.4.1 Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d372de65",
   "metadata": {},
   "source": [
    "Hyperparameters are tuned using 5-fold cross validation. Due to time constraints, parameters are tuned in two steps rather than in 1 go. Random Search CV can also be used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "5c19ea5d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'min_samples_leaf': 5, 'n_estimators': 500}\n",
      "Best score is 0.7249112697708067\n"
     ]
    }
   ],
   "source": [
    "# Number of trees in ensemble, default = 100\n",
    "n_estimators = [100, 300, 500, 1000]\n",
    "\n",
    "# Min samples in leaf, default = 1\n",
    "min_samples_leaf = [1, 5, 25, 50]\n",
    "param_grid = {'n_estimators': n_estimators,\n",
    "             'min_samples_leaf': min_samples_leaf }\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = RandomForestClassifier(oob_score = True, n_jobs = -1, random_state =42)\n",
    "rfc_cv = GridSearchCV(model, param_grid, cv=5)\n",
    "rfc_cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(rfc_cv.best_params_)) \n",
    "print(\"Best score is {}\".format(rfc_cv.best_score_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "72ba6a33",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'max_depth': 15, 'min_samples_split': 25}\n",
      "Best score is 0.7243951571799678\n"
     ]
    }
   ],
   "source": [
    "# Fix min_samples_leaf and n_estimators and tune others\n",
    "max_depth = [1, 3, 6, 10, 15] #default = 0\n",
    "min_samples_split = [25, 100, 150, 200] #default = 2\n",
    "\n",
    "model = RandomForestClassifier(oob_score = True, \n",
    "                               n_jobs = -1, \n",
    "                               random_state = 42,\n",
    "                              min_samples_leaf = 5,\n",
    "                              n_estimators = 500)\n",
    "\n",
    "param_grid = {'max_depth': max_depth, #default is 6\n",
    "             'min_samples_split': min_samples_split, #default is 100\n",
    "             }\n",
    "\n",
    "rfc_cv = GridSearchCV(model, param_grid, cv=5)\n",
    "rfc_cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(rfc_cv.best_params_)) \n",
    "print(\"Best score is {}\".format(rfc_cv.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5cf080b5",
   "metadata": {},
   "source": [
    "#### 2.4.2 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0be60e87",
   "metadata": {},
   "source": [
    "Pass the best parameters into Random Forest Classifier and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "258e8a11",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Metrics\n",
      "Random Forest Plain Accuracy:72.78%\n",
      "TPR:60.74%\n",
      "FNR:39.26%\n",
      "TNR:81.48%\n",
      "FPR:18.52%\n",
      "Recall:60.74%\n",
      "Precision:70.33%\n",
      "AUC Score:0.711\n",
      "\n",
      "Train Set Metrics\n",
      "Random Forest Plain Accuracy:78.98%\n",
      "Recall:69.08%\n",
      "Precision:78.26%\n",
      "AUC Score:0.776\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# p (number of features) is set at default as there is little difference in the various methods\n",
    "\n",
    "rfc = RandomForestClassifier(random_state = 42, \n",
    "                             max_features = 'sqrt', \n",
    "                             min_samples_leaf = 5, \n",
    "                             n_estimators = 500,\n",
    "                            max_depth = 15,\n",
    "                            min_samples_split = 25)\n",
    "rfc.fit(X_train, y_train)\n",
    "\n",
    "classification_summary(X_train = X_train, \n",
    "                       X_test = X_test,\n",
    "                       y_train = y_train, \n",
    "                       y_test = y_test, \n",
    "                       model_name = rfc, \n",
    "                       model_name_text = 'Random Forest')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fb7a1692",
   "metadata": {},
   "source": [
    "### 2.5 XGBoost Classifier<a class=\"anchor\" id=\"section2.5\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "70a08346",
   "metadata": {},
   "source": [
    "#### 2.5.1 Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ab37b8cb",
   "metadata": {},
   "source": [
    "Hyperparameters are tuned using 5-fold cross validation. Due to time constraints, parameters are tuned in two steps rather than in 1 go. Random Search CV can also be used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "eccb219b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'max_depth': 3, 'n_estimators': 100}\n",
      "Best score is 0.7461577554174477\n"
     ]
    }
   ],
   "source": [
    "max_depth = [3, 6, 10]\n",
    "n_estimators = [100, 300, 500]\n",
    "\n",
    "param_grid = {'max_depth': max_depth, #default is 6\n",
    "             'n_estimators': n_estimators, #default is 100\n",
    "             }\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = XGBClassifier(use_label_encoder=False, random_state = 42, \n",
    "                      eval_metric='auc',\n",
    "                     objective = 'binary:logistic')\n",
    "\n",
    "cv = GridSearchCV(model, param_grid, cv=5, n_jobs = -1)\n",
    "cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(cv.best_params_)) \n",
    "print(\"Best score is {}\".format(cv.best_score_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "84b26a59",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'colsample_bytree': 0.7, 'gamma': 0.1}\n",
      "Best score is 0.7484230377166157\n"
     ]
    }
   ],
   "source": [
    "# Fix max_depth and n_estimators and tune others\n",
    "colsample_bytree = [0.1, 0.3, 0.7, 1] #default = 0.3\n",
    "gamma = [0.1, 0.3, 0.7, 1] #default = 0\n",
    "\n",
    "param_grid = {'colsample_bytree': colsample_bytree, #default is 6\n",
    "             'gamma': gamma, #default is 100\n",
    "             }\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = XGBClassifier(use_label_encoder=False, random_state = 42, \n",
    "                      eval_metric='auc',\n",
    "                     objective = 'binary:logistic',\n",
    "                     max_depth = 3,\n",
    "                     n_estimators = 100)\n",
    "\n",
    "cv = GridSearchCV(model, param_grid, cv=5, n_jobs = -1)\n",
    "cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(cv.best_params_)) \n",
    "print(\"Best score is {}\".format(cv.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d676f3af",
   "metadata": {},
   "source": [
    "#### 2.5.2 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9330fc60",
   "metadata": {},
   "source": [
    "Pass the best parameters into XGBoost and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "6cc24e0e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Metrics\n",
      "XGBoost Plain Accuracy:74.92%\n",
      "TPR:68.67%\n",
      "FNR:31.33%\n",
      "TNR:79.44%\n",
      "FPR:20.56%\n",
      "Recall:68.67%\n",
      "Precision:70.71%\n",
      "AUC Score:0.741\n",
      "\n",
      "Train Set Metrics\n",
      "XGBoost Plain Accuracy:75.93%\n",
      "Recall:69.35%\n",
      "Precision:72.18%\n",
      "AUC Score:0.75\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "xgb = XGBClassifier(use_label_encoder=False, random_state = 42, \n",
    "                    eval_metric='auc',\n",
    "                    objective = 'binary:logistic',\n",
    "                    max_depth = 3,\n",
    "                    n_estimators = 100, \n",
    "                    colsample_bytree = 0.7, \n",
    "                    gamma = 0.1)\n",
    "\n",
    "xgb.fit(X_train, y_train)\n",
    "\n",
    "classification_summary(X_train = X_train, \n",
    "                       X_test = X_test,\n",
    "                       y_train = y_train, \n",
    "                       y_test = y_test, \n",
    "                       model_name = xgb, \n",
    "                       model_name_text = 'XGBoost')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc574886",
   "metadata": {},
   "source": [
    "### 2.6 Classifier Evaluation & Selection<a class=\"anchor\" id=\"section2.6\"></a>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "c3e18fec",
   "metadata": {},
   "source": [
    "#### Best Performance: XGBoost Classifier\n",
    "\n",
    "Overall, we should proceeed with XGBoost Classifier to predict offer redemption given as it has the highest precision, recall, AUC and accuracy by a long shot.\n",
    "\n",
    "<b>Key Observations</b>\n",
    "\n",
    "- From the Exploratory Data Analysis (EDA) is known that only 42% of all offers sent out are successfully seen & completed\n",
    "- XGBoost (and arguable any classifier) will be able to improve upon this significantly, to 71%, represeting <b>+29% improvement</b> (from the precision metric) \n",
    "- Logistic Regression generally performed the worst, strongly suggesting that a linear decision boundary may not be the best fit for offer sucess. \n",
    "\n",
    "Each classifiers' performance is summarised below, with best performance in green and worst in red. Test Set metrics are what are used for evaluation and Train Set is included for reference only.\n",
    "\n",
    "<img src = \"./images/classifier_results.png\" width=\"600\"/>\n",
    "\n",
    "The ROC Curve further reinforces this as Random Forest and XGBoost visibly have the maximum area under curve (AUC).\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "90bbb4d5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "dict_classifiers = {\n",
    "    \"Logistic Regression\": logreg,\n",
    "    \"SVM\": svm,\n",
    "    \"RandomForest\": rfc,\n",
    "    \"XGBoost\": xgb\n",
    "}\n",
    "\n",
    "plt.clf()\n",
    "\n",
    "fig, ax = plt.subplots(figsize=[8,6])\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.title('ROC Curve', fontsize = 20)\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "\n",
    "lines = []\n",
    "\n",
    "for i, (model, model_instantiation) in enumerate(dict_classifiers.items()):\n",
    "    # Compute predicted probabilities: y_pred_prob\n",
    "    \n",
    "    if model == 'SVM':\n",
    "        y_pred_prob = model_instantiation.predict_proba(X_test_scaled)[:,1]\n",
    "    \n",
    "    else:\n",
    "        y_pred_prob = model_instantiation.predict_proba(X_test)[:,1]\n",
    "\n",
    "    # Generate ROC curve values: fpr, tpr, thresholds\n",
    "    fpr, tpr, thresholds = roc_curve(y_test, y_pred_prob)\n",
    "    \n",
    "    # Plot ROC curve\n",
    "    # plt.plot(fpr, tpr)\n",
    "    lines += plt.plot(fpr,tpr, label=\"{}\".format(model))\n",
    "\n",
    "# Create manual labelling\n",
    "labels = [l.get_label() for l in lines]\n",
    "plt.legend(lines, labels)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7541c409",
   "metadata": {},
   "source": [
    "### 2.7 Findings: The Most Important Features Predicting Offer Success<a class=\"anchor\" id=\"section2.7\"></a>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "95374607",
   "metadata": {},
   "source": [
    "With opaque classifiers like XGBoost, it is useful to use <b>permutation feature imporance</b> to see which features are most impactful. \n",
    "\n",
    "In a nutshell, this metric tells us which features, if present, are most likely to result in a better classification outcome. Refer to sklearn's insightful summary on [permutation feature importance](https://scikit-learn.org/stable/modules/permutation_importance.html)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "3fbe4b07",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "result = permutation_importance(\n",
    "    xgb, X_test, y_test, n_repeats=10, random_state=42, n_jobs=-1\n",
    ")\n",
    "\n",
    "feature_names = [f\"{i}\" for i in X.columns]\n",
    "importances = result.importances_mean\n",
    "std = result.importances_std\n",
    "\n",
    "my_dict = {'feature_names': feature_names, 'feature_impt': importances, 'std': std}\n",
    "\n",
    "pi = pd.DataFrame.from_dict(my_dict)\n",
    "pi.sort_values(by = 'feature_impt', ascending = False, inplace = True)\n",
    "\n",
    "# Visualize feature importances\n",
    "fig, ax = plt.subplots(figsize=[8,6])\n",
    "plt.barh(y = pi['feature_names'], width = pi['feature_impt'], xerr=pi['std'])\n",
    "ax.set_title(\"Feature importances using permutation\", fontsize = 20)\n",
    "ax.set_xlabel(\"Mean accuracy decrease\")\n",
    "ax.invert_yaxis()  # labels read top-to-bottom\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "00f11442",
   "metadata": {},
   "source": [
    "#### The 3 Most Important Features Predicting Offer Completion\n",
    "\n",
    "We can clearly see that the top 3 most important factors in whether or not a promotion is redeemed are *days*, *member_duration* and *channel_social* (social media targeting).\n",
    "\n",
    "This is valuable information from a marketing perspective as one may reasonably conclude that:\n",
    "1. <b>Long-time members may be more likely to redeem promotions</b> and should be more heavily targeted, whereas members in other age groups may need to be engaged in other ways.\n",
    "2. <b>Sharing the promotion across social media is important</b>, suggesting that other distribution channels (website/ mobile app) were not so effective.\n",
    "\n",
    "As for the importance of *days*, this may be because offers are sent in bursts through the month. Hence, the closer the date is to those bursts, the higher likelihood of offer success. It may be that as time elapses without an offer completion, customers are more likely to have seen but are ignoring the offer. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7086dc92",
   "metadata": {},
   "source": [
    "## 3. Predicting Amount Spent through Regression<a class=\"anchor\" id=\"section3\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14753aca",
   "metadata": {},
   "source": [
    "#### Target Variable: *amount* (binary)\n",
    "\n",
    "In this section, we train and select a regressor that helps to predict amount spent <b>given offer success</b>. This will help to assign an expected spend value to each expected offer redemption.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d0fcaf81",
   "metadata": {},
   "source": [
    "### 3.1 Create Train-Test Split and Helper Functions<a class=\"anchor\" id=\"section3.1\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3e9d0303",
   "metadata": {},
   "source": [
    "30% of data is withheld as the test set. Only data where an offer has been successful is used for model training as we are only concerned with the conditional amount spent given a succesful offer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "a05c7add",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Select only success scenarios\n",
    "df_reg = df_all[df_all['success']==1]\n",
    "\n",
    "y_reg = df_reg['amount']\n",
    "\n",
    "features = ['age', 'income', 'member_duration','reward', 'difficulty', \n",
    "        'days', 'channel_web', 'channel_mobile', \n",
    "        'channel_social', 'gender_dum_M', 'offer_type_dum_discount']\n",
    "\n",
    "X_reg = df_reg[features]\n",
    "\n",
    "# Define train-test split\n",
    "X_train_reg, X_test_reg, y_train_reg, y_test_reg = train_test_split(X_reg, y_reg, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad882518",
   "metadata": {},
   "source": [
    "A helper function is also created to easily print regression scores, using mean squared error (MSE) as the success metric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "67580f77",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Helper function to print scores\n",
    "\n",
    "def regression_scores(model, X_train, X_test, y_train, y_test):\n",
    "    \n",
    "    pred_train= model.predict(X_train)\n",
    "    pred_test= model.predict(X_test)\n",
    "    \n",
    "    train_MSE = mean_squared_error(y_train, pred_train)\n",
    "    test_MSE = mean_squared_error(y_test, pred_test)\n",
    "    \n",
    "    print('R2 train', round(model.score(X_train, y_train),2))\n",
    "    print('Train MSE:', round(train_MSE,2))\n",
    "    print('')\n",
    "    print('R2 test', round(model.score(X_test, y_test),2))\n",
    "    print('Test MSE:', round(test_MSE,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "087fe4b2",
   "metadata": {},
   "source": [
    "### 3.2 Linear Regression<a class=\"anchor\" id=\"section3.2\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d9e5cd28",
   "metadata": {},
   "source": [
    "Fit a simple linear regression to get a baseline."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "18777861",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 train 0.46\n",
      "Train MSE: 39.42\n",
      "\n",
      "R2 test 0.47\n",
      "Test MSE: 39.05\n"
     ]
    }
   ],
   "source": [
    "lr = LinearRegression()\n",
    "lr.fit(X_train_reg,y_train_reg)\n",
    "\n",
    "regression_scores(lr, X_train_reg, X_test_reg, y_train_reg, y_test_reg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ba91c4bd",
   "metadata": {},
   "source": [
    "### 3.3 Random Forest Regressor<a class=\"anchor\" id=\"section3.3\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd43bde0",
   "metadata": {},
   "source": [
    "#### 3.3.1 Hyperparameter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "62c7135b",
   "metadata": {},
   "source": [
    "Hyperparameters are tuned using 5-fold cross validation. Due to time constraints, parameters are tuned in two steps rather than in 1 go. Random Search CV can also be used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "792f0299",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'min_samples_leaf': 25, 'n_estimators': 1000}\n",
      "Best score is 0.2835803732150929\n"
     ]
    }
   ],
   "source": [
    "# Number of trees in ensemble, default = 100\n",
    "n_estimators = [100, 300, 500, 1000]\n",
    "\n",
    "# Min samples in leaf, default = 1\n",
    "min_samples_leaf = [25, 50, 100, 200]\n",
    "param_grid = {'n_estimators': n_estimators,\n",
    "             'min_samples_leaf': min_samples_leaf}\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = RandomForestRegressor(n_jobs = -1, random_state =42)\n",
    "rfr_cv = GridSearchCV(model, param_grid, cv=5)\n",
    "rfr_cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(rfr_cv.best_params_)) \n",
    "print(\"Best score is {}\".format(rfr_cv.best_score_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "5a9e703c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tuned Parameters: {'max_depth': 15}\n",
      "Best score is 0.28231903186834667\n"
     ]
    }
   ],
   "source": [
    "# Fix n_estimators and min_samples_leaf and tune depth\n",
    "m_depth = [1,3,6,10,15]\n",
    "\n",
    "param_grid = {'max_depth': m_depth}\n",
    "\n",
    "# Instantiate the GridSearchCV object\n",
    "model = RandomForestRegressor(n_jobs = -1, random_state =42,min_samples_leaf = 25, n_estimators = 1000)\n",
    "rfr_cv = GridSearchCV(model, param_grid, cv=5)\n",
    "rfr_cv.fit(X_train,y_train)\n",
    "\n",
    "# Print the tuned parameters and score\n",
    "print(\"Tuned Parameters: {}\".format(rfr_cv.best_params_)) \n",
    "print(\"Best score is {}\".format(rfr_cv.best_score_))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9ead1cb7",
   "metadata": {},
   "source": [
    "#### 3.3.2 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a59e7524",
   "metadata": {},
   "source": [
    "Pass the best parameters into Random Forest and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "d18bb28b",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 train 0.57\n",
      "Train MSE: 31.73\n",
      "\n",
      "R2 test 0.5\n",
      "Test MSE: 36.51\n"
     ]
    }
   ],
   "source": [
    "rfr = RandomForestRegressor(n_jobs = -1, random_state =42, \n",
    "                            min_samples_leaf = 25, \n",
    "                            n_estimators = 1000,\n",
    "                           max_depth = 15)\n",
    "\n",
    "rfr.fit(X_train_reg,y_train_reg)\n",
    "\n",
    "regression_scores(rfr, X_train_reg, X_test_reg, y_train_reg, y_test_reg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e94336d2",
   "metadata": {},
   "source": [
    "### 3.4 K-Nearest Neighbours Regressor<a class=\"anchor\" id=\"section3.4\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eafeecda",
   "metadata": {},
   "source": [
    "#### 3.4.1 Selection of Optimal K"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af40a4cd",
   "metadata": {},
   "source": [
    "Using the elbow method, we visually select <b>K = 15</b> as the point where gain in accuracy from that point onwards is marginal."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "34b94da9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "error = []\n",
    "\n",
    "for i in range(1,40):\n",
    "    \n",
    "    knn = KNeighborsRegressor(n_neighbors=i)\n",
    "    knn.fit(X_train,y_train)\n",
    "    pred_i = knn.predict(X_test)\n",
    "    error.append(mean_squared_error(y_test, pred_i))\n",
    "    \n",
    "# Visualise error rate to see where it tapers off\n",
    "plt.figure(figsize=(10,6))\n",
    "plt.plot(range(1,40),error,color='blue', linestyle='dashed', marker='o',\n",
    "         markerfacecolor='red', markersize=10)\n",
    "plt.title('Error Rate (MSE) vs. K Value', fontsize = 20)\n",
    "plt.xlabel('K')\n",
    "plt.ylabel('Error Rate')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9d53a3c",
   "metadata": {},
   "source": [
    "#### 3.4.2 Train Model & Inspect Performance on Test Set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ca29e5d2",
   "metadata": {},
   "source": [
    "Pass the best parameters into KNN Regressor and obtain test score. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "d421ebcc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 train 0.51\n",
      "Train MSE: 36.14\n",
      "\n",
      "R2 test 0.44\n",
      "Test MSE: 41.34\n"
     ]
    }
   ],
   "source": [
    "# Running with optimized k\n",
    "knn = KNeighborsRegressor(n_neighbors=15)\n",
    "knn.fit(X_train_reg,y_train_reg)\n",
    "\n",
    "regression_scores(knn, X_train_reg, X_test_reg, y_train_reg, y_test_reg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1f6021b3",
   "metadata": {},
   "source": [
    "### 3.5 Regressor Evaluation & Selection<a class=\"anchor\" id=\"section3.5\"></a>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "3ca891b5",
   "metadata": {},
   "source": [
    "#### Best Performance: Random Forest Regressor\n",
    "\n",
    "Overall, we should proceeed with Random Forest Regressor to predict offer redemption given as it has the highest overall $r^2$ and lowest error.\n",
    "\n",
    "Each regressors' performance is summarised below, with best performance in green and worst in red. Test Set metrics are what are used for evaluation and Train Set is included for reference only.\n",
    "\n",
    "<img src = \"./images/regressor_results.png\" width=\"600\"/>\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6818bd49",
   "metadata": {},
   "source": [
    "### 3.6 Findings: The Most Important Features Predicting Amount Spent<a class=\"anchor\" id=\"section3.6\"></a>"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "a8a865b7",
   "metadata": {},
   "source": [
    "With opaque regressors like RandomForest, it is useful to use <b>permutation feature imporance</b> to see which features are most impactful. \n",
    "\n",
    "In a nutshell, this metric tells us which features, if present, are most likely to result in a better classification outcome. Refer to sklearn's insightful summary on [permutation feature importance](https://scikit-learn.org/stable/modules/permutation_importance.html)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "be33445d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "result = permutation_importance(\n",
    "    rfr, X_test_reg, y_test_reg, n_repeats=10, random_state=42, n_jobs=-1\n",
    ")\n",
    "\n",
    "feature_names = [f\"{i}\" for i in X.columns]\n",
    "importances = result.importances_mean\n",
    "std = result.importances_std\n",
    "\n",
    "my_dict = {'feature_names': feature_names, 'feature_impt': importances, 'std': std}\n",
    "\n",
    "pi = pd.DataFrame.from_dict(my_dict)\n",
    "pi.sort_values(by = 'feature_impt', ascending = False, inplace = True)\n",
    "\n",
    "# Visualize feature importances\n",
    "fig, ax = plt.subplots(figsize=[8,6])\n",
    "plt.barh(y = pi['feature_names'], width = pi['feature_impt'], xerr=pi['std'])\n",
    "ax.set_title(\"Feature importances using permutation\", fontsize = 20)\n",
    "ax.set_xlabel(\"Mean accuracy decrease\")\n",
    "ax.invert_yaxis()  # labels read top-to-bottom\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6d530933",
   "metadata": {},
   "source": [
    "#### The Most Important Feature Predicting Offer Completion\n",
    "\n",
    "We can clearly see that the most important factor in determining amount spent is the individual's income. This has two important implications:\n",
    "\n",
    "1. This underscores the importance of good data collection of the customer income feature, where legal\n",
    "2. From a marketing perspective, this means <b>more targeted customer acquisition campaigns to high-income individuals</b> would be justified as they eventually will also spend more"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e0e6be8",
   "metadata": {},
   "source": [
    "## 4. Combined Expected Value Model<a class=\"anchor\" id=\"section4\"></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "21bb2304",
   "metadata": {},
   "source": [
    "Following this we can apply both models to optimise the promotions sent to a specific customer.\n",
    "\n",
    "For example, we can start with person id *ffe5257abf8840b395e1ee6b29894637* who is a 77 year old male, with 75,000 income and has been a member for 338 days. In the experiment, he received 5 promotions in a month of which only 1 was successful. He spent \\$15.22 as a result of that successful buy-one-get-one offer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "f0524f0e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>promo_name</th>\n",
       "      <th>success</th>\n",
       "      <th>amount</th>\n",
       "      <th>age</th>\n",
       "      <th>income</th>\n",
       "      <th>member_duration</th>\n",
       "      <th>reward</th>\n",
       "      <th>difficulty</th>\n",
       "      <th>days</th>\n",
       "      <th>channel_web</th>\n",
       "      <th>channel_mobile</th>\n",
       "      <th>channel_social</th>\n",
       "      <th>gender_dum_M</th>\n",
       "      <th>offer_type_dum_discount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>6172</th>\n",
       "      <td>disc-1</td>\n",
       "      <td>0</td>\n",
       "      <td>34.95</td>\n",
       "      <td>77</td>\n",
       "      <td>75000.0</td>\n",
       "      <td>338</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>9.00</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12334</th>\n",
       "      <td>bogo-1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>77</td>\n",
       "      <td>75000.0</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>17.75</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43523</th>\n",
       "      <td>bogo-2</td>\n",
       "      <td>1</td>\n",
       "      <td>15.22</td>\n",
       "      <td>77</td>\n",
       "      <td>75000.0</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>15.50</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49821</th>\n",
       "      <td>disc-4</td>\n",
       "      <td>0</td>\n",
       "      <td>34.95</td>\n",
       "      <td>77</td>\n",
       "      <td>75000.0</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>20</td>\n",
       "      <td>9.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49822</th>\n",
       "      <td>disc-4</td>\n",
       "      <td>0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>77</td>\n",
       "      <td>75000.0</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>20</td>\n",
       "      <td>21.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      promo_name  success  amount  age   income  member_duration  reward  \\\n",
       "6172      disc-1        0   34.95   77  75000.0              338       2   \n",
       "12334     bogo-1        0    0.00   77  75000.0              338       5   \n",
       "43523     bogo-2        1   15.22   77  75000.0              338       5   \n",
       "49821     disc-4        0   34.95   77  75000.0              338       5   \n",
       "49822     disc-4        0    0.00   77  75000.0              338       5   \n",
       "\n",
       "       difficulty   days  channel_web  channel_mobile  channel_social  \\\n",
       "6172           10   9.00            1               1               0   \n",
       "12334           5  17.75            1               1               1   \n",
       "43523           5  15.50            1               1               0   \n",
       "49821          20   9.00            1               0               0   \n",
       "49822          20  21.00            1               0               0   \n",
       "\n",
       "       gender_dum_M  offer_type_dum_discount  \n",
       "6172              1                        1  \n",
       "12334             1                        0  \n",
       "43523             1                        0  \n",
       "49821             1                        1  \n",
       "49822             1                        1  "
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_all[df_all['person'] == 'ffe5257abf8840b395e1ee6b29894637'][['promo_name','success','amount']+features]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75f4ced3",
   "metadata": {},
   "source": [
    "### 4.1 Offer Ranking Using Trained Models<a class=\"anchor\" id=\"section4.1\"></a>\n",
    "\n",
    "<b> Steps </b>\n",
    "1. Demographic and promotions data first fed into the classifier to get $ P(OfferSuccess)$\n",
    "2. The same features are then fed to the regressor to get $Expected Spend$\n",
    "3. The promotions are then ranked by Expected Value ($P(OfferSuccess) * Expected Spend$)\n",
    "\n",
    "<b> Assumptions </b>\n",
    "- We should only serve the top 4 best promotions in a single month\n",
    "    - This prevents brand equity dilution by restricting offers to just once per week\n",
    "    - This preserves the mode number of offers sent in the current dataset (ranges from 1 to 10)\n",
    "- We should not repeat promotions\n",
    "    - Offer success probabilities may change for subsequent promotions of the same type, this has not been modelled here"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "6fb23344",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>promo_name</th>\n",
       "      <th>age</th>\n",
       "      <th>income</th>\n",
       "      <th>member_duration</th>\n",
       "      <th>reward</th>\n",
       "      <th>difficulty</th>\n",
       "      <th>days</th>\n",
       "      <th>channel_web</th>\n",
       "      <th>channel_mobile</th>\n",
       "      <th>channel_social</th>\n",
       "      <th>gender_dum_M</th>\n",
       "      <th>offer_type_dum_discount</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>bogo-4</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>10</td>\n",
       "      <td>10</td>\n",
       "      <td>15</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>bogo-3</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>10</td>\n",
       "      <td>10</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>bogo-2</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>disc-4</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>20</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>disc-3</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>3</td>\n",
       "      <td>7</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>disc-2</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>bogo-1</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>disc-1</td>\n",
       "      <td>77</td>\n",
       "      <td>75000</td>\n",
       "      <td>338</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>15</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  promo_name  age  income  member_duration  reward  difficulty  days  \\\n",
       "0     bogo-4   77   75000              338      10          10    15   \n",
       "1     bogo-3   77   75000              338      10          10    15   \n",
       "2     bogo-2   77   75000              338       5           5    15   \n",
       "3     disc-4   77   75000              338       5          20    15   \n",
       "4     disc-3   77   75000              338       3           7    15   \n",
       "5     disc-2   77   75000              338       2          10    15   \n",
       "6     bogo-1   77   75000              338       5           5    15   \n",
       "7     disc-1   77   75000              338       2          10    15   \n",
       "\n",
       "   channel_web  channel_mobile  channel_social  gender_dum_M  \\\n",
       "0            0               1               1             1   \n",
       "1            1               1               1             1   \n",
       "2            1               1               0             1   \n",
       "3            1               0               0             1   \n",
       "4            1               1               1             1   \n",
       "5            1               1               1             1   \n",
       "6            1               1               1             1   \n",
       "7            1               1               0             1   \n",
       "\n",
       "   offer_type_dum_discount  \n",
       "0                        0  \n",
       "1                        0  \n",
       "2                        0  \n",
       "3                        1  \n",
       "4                        1  \n",
       "5                        1  \n",
       "6                        0  \n",
       "7                        1  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example = pd.read_csv('data/example.csv')\n",
    "example"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "96e5f190",
   "metadata": {},
   "source": [
    "<b>Step 1:</b> Demographic and promotions data first fed into the classifier to get  𝑃(𝑂𝑓𝑓𝑒𝑟𝑆𝑢𝑐𝑐𝑒𝑠𝑠)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "b7302225",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.41, 0.46, 0.34, 0.27, 0.63, 0.61, 0.49, 0.33], dtype=float32)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "float_formatter = \"{:.2f}\".format\n",
    "np.set_printoptions(formatter={'float_kind':float_formatter})\n",
    "\n",
    "example_p_offer_success = xgb.predict_proba(example.iloc[:,1:])# Exclude promo_name from features \n",
    "example_p_offer_success = example_p_offer_success[:,1] # Take only the probability of class 1 i.e. offer success = True \n",
    "example_p_offer_success"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f4bc92ef",
   "metadata": {},
   "source": [
    "<b>Step 2:</b> The same features are then fed to the regressor to get  $𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑆𝑝𝑒𝑛𝑑$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "d467198e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([19.74, 19.63, 19.63, 20.40, 19.33, 19.43, 18.82, 19.98])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example_exp_spend = rfr.predict(example.iloc[:,1:])# Exclude promo_name from features \n",
    "example_exp_spend"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7dd9d65c",
   "metadata": {},
   "source": [
    "<b>Step 3:</b> The promotions are then ranked by Expected Value ($P(OfferSuccess) * Expected Spend$)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "d35ac83f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>promo_name</th>\n",
       "      <th>P(Offer Success)</th>\n",
       "      <th>Expected Spend</th>\n",
       "      <th>P(Offer Success) x Expected Spend</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>bogo-4</td>\n",
       "      <td>0.406683</td>\n",
       "      <td>19.743503</td>\n",
       "      <td>8.029352</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>bogo-3</td>\n",
       "      <td>0.461236</td>\n",
       "      <td>19.629418</td>\n",
       "      <td>9.053791</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>bogo-2</td>\n",
       "      <td>0.339964</td>\n",
       "      <td>19.631866</td>\n",
       "      <td>6.674131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>disc-4</td>\n",
       "      <td>0.265187</td>\n",
       "      <td>20.404650</td>\n",
       "      <td>5.411046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>disc-3</td>\n",
       "      <td>0.628778</td>\n",
       "      <td>19.328527</td>\n",
       "      <td>12.153360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>disc-2</td>\n",
       "      <td>0.608454</td>\n",
       "      <td>19.434354</td>\n",
       "      <td>11.824909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>bogo-1</td>\n",
       "      <td>0.494991</td>\n",
       "      <td>18.822586</td>\n",
       "      <td>9.317015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>disc-1</td>\n",
       "      <td>0.333459</td>\n",
       "      <td>19.984149</td>\n",
       "      <td>6.663903</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  promo_name  P(Offer Success)  Expected Spend  \\\n",
       "0     bogo-4          0.406683       19.743503   \n",
       "1     bogo-3          0.461236       19.629418   \n",
       "2     bogo-2          0.339964       19.631866   \n",
       "3     disc-4          0.265187       20.404650   \n",
       "4     disc-3          0.628778       19.328527   \n",
       "5     disc-2          0.608454       19.434354   \n",
       "6     bogo-1          0.494991       18.822586   \n",
       "7     disc-1          0.333459       19.984149   \n",
       "\n",
       "   P(Offer Success) x Expected Spend  \n",
       "0                           8.029352  \n",
       "1                           9.053791  \n",
       "2                           6.674131  \n",
       "3                           5.411046  \n",
       "4                          12.153360  \n",
       "5                          11.824909  \n",
       "6                           9.317015  \n",
       "7                           6.663903  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ranking = example[['promo_name']]\n",
    "ranking['P(Offer Success)'] = example_p_offer_success\n",
    "ranking['Expected Spend'] = example_exp_spend\n",
    "ranking['P(Offer Success) x Expected Spend'] = ranking['P(Offer Success)'] * ranking['Expected Spend']\n",
    "ranking"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bbf0ebb8",
   "metadata": {},
   "source": [
    "<b>Final Result</b>: The top 4 offers to be served are as disc-3, disc-2, bogo-1 and bogo-3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "5b03022d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>promo_name</th>\n",
       "      <th>P(Offer Success)</th>\n",
       "      <th>Expected Spend</th>\n",
       "      <th>P(Offer Success) x Expected Spend</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>disc-3</td>\n",
       "      <td>0.628778</td>\n",
       "      <td>19.328527</td>\n",
       "      <td>12.153360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>disc-2</td>\n",
       "      <td>0.608454</td>\n",
       "      <td>19.434354</td>\n",
       "      <td>11.824909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>bogo-1</td>\n",
       "      <td>0.494991</td>\n",
       "      <td>18.822586</td>\n",
       "      <td>9.317015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>bogo-3</td>\n",
       "      <td>0.461236</td>\n",
       "      <td>19.629418</td>\n",
       "      <td>9.053791</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  promo_name  P(Offer Success)  Expected Spend  \\\n",
       "4     disc-3          0.628778       19.328527   \n",
       "5     disc-2          0.608454       19.434354   \n",
       "6     bogo-1          0.494991       18.822586   \n",
       "1     bogo-3          0.461236       19.629418   \n",
       "\n",
       "   P(Offer Success) x Expected Spend  \n",
       "4                          12.153360  \n",
       "5                          11.824909  \n",
       "6                           9.317015  \n",
       "1                           9.053791  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sort promo_name by score\n",
    "ranking.sort_values('P(Offer Success) x Expected Spend', ascending = False).iloc[0:4]"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}